This paper considers the problem of leaderless self-tuning consensus in multiagent discrete time systems with unknown control directions. We introduce novel algorithms for adaptive tuning of interagent coupling parameters. The control algorithm is fully distributed and the agent control signals are continuous functions in their arguments. Networked systems on directed graphs are analyzed. Assuming that the graph is strongly connected it is proved that all agent states converge toward the same value, and the interagent coupling parameters are convergent sequences.